Cardiovascular diseases have become one of the leading causes of death, and often lead to other medical conditions such as strokes, hypertension, heart failure and arrhythmia. In the field of biomedical engineering, automatic abnormal heart sound detection can be considered a major prior step to cardiovascular disease diagnosis. The process of identifying whether a given heart sound is normal or abnormal can be divided into three major steps: segmentation, feature extraction, and classification . Firstly, the segmentation technique locates the fundamental heart sounds of the Phonocardiogram (PCG) signal: S1 (first heart sound) and S2 (second heart sound);
Traditionally, abnormal heart sound classification is framed as a three-stage process. The first stage involves segmenting the phonocardiogram to detect fundamental heart sounds; after which features are extracted and classification is performed. Some researchers in the field argue the segmentation step is an unwanted computational burden, whereas others embrace it as a prior step to feature extraction. When comparing accuracies achieved by studies that have segmented heart sounds before analysis with those who have overlooked that step, the question of whether to segment heart sounds before feature extraction is still open. In this study, we explicitly examine the importance of heart sound segmentation as a prior step for heart sound classification, and then seek to apply the obtained insights to propose a robust classifier for abnormal heart sound detection. Furthermore, recognizing the pressing need for explainable Artificial Intelligence (AI) models in the medical domain, we also unveil hidden representations learned by the classifier using model interpretation techniques.
In their extensive review of heart sound analysis research from the years 2015 to 2017. Moreover, according to the recent study by Zhang et al. , which introduces the best-performing model to date, segmentation is not a mandatory step for heart anomaly detection because the primary focus of the algorithm should be detecting the presence of an anomaly, which can be achieved without locating it. However, examining the recent literature, studies that considered segmentation as a prior step have also achieved competitive performance. Therefore further investigations are required.
We explicitly examine the importance of heart sound segmentation as a prior step for heart sound classification, and then seek to apply the obtained insights to propose a robust classifier for abnormal heart sound detection. Furthermore, recognizing the pressing need for explainable Artificial Intelligence (AI) models in the medical domain, we also unveil hidden representations learned by the classifier using model interpretation techniques.
- Operating system : Windows.
- Coding Language : Python.
System : Pentium IV 2.4 GHz or intel
Hard Disk : 40 GB.
Floppy Drive : 1.44 Mb.
Mouse : Optical Mouse.
Ram : 512 Mb.
The main objective of this research was to discover the advantages of heart sound segmentation for abnormal heart sound classification, and derive a robust explainable model using those insights. Considering the models proposed in this study, Model 3 from Experiment I and the model described in Experiment II, have superior and stable performance for normal-abnormal heart sound classification compared to the state-of-the-art. The final model, which is designed by considering the conclusions made in the first experiment, outperforms the best performing classifier in the literature with an accuracy by 5%. Furthermore, the architecture of the classifier only has three convolution layers and three fully connected layers, making it a simple but efficient model.